Kyle Chan
@kyleichan • 48,640 subscribers
Research Fellow at @BrookingsInst. China's tech & industrial policy: AI, chips, robotics, EVs, clean energy. Newsletter + podcast: https://t.co/D6k0b2dJgZ
Videos
This Chinese humanoid robot just shattered the world record for a half marathon, finishing in 50 min 26 sec. This video shows its crash just meters before the finish line where it had to be picked up by a team of humans. The robot is from Honor, the smartphone maker and Huawei spin-off. This robot was teleoperated while others were autonomous. It seems like all the robots had battery swaps along the way.
Kyle Chan5,292,668 views • 3 months ago
In 1978, Deng Xiaoping rode a Japanese bullet train for the first time. 3 decades earlier, Japan was China’s greatest enemy. In the 1970s, China began to open up & learn from others, incl. Japan. 3 decades later, China began to build the world’s largest high-speed rail network.
Kyle Chan753,883 views • 11 months ago
Unitree founder Wang Xingxing: In robotics, locomotion and basic motion is mostly solved. But grasping and manipulation—anything related to haptics—hasn’t been solved. That’s the key bottleneck preventing them from being deployed at scale in factories and homes. He says that simulation is much faster for training but for manipulation tasks you still need real-world training data—for now.
Kyle Chan67,761 views • 3 months ago
US Treasury Secretary Scott Bessent on AI: “What we’ve had in the past month was a step change in the power of one large language model, but we’re going to see it from the other AI companies. And it’s important, Maria, that the U.S. stays ahead here. Imagine if China or some non-state actor were ahead of us. So what we’re determined to do is work with our AI companies to allow them to continue to innovate, but our charge in the U.S. government is maintaining safety. And there is a very important calculus here between innovation and safety. And the U.S. government, we’re going to make sure that things stay safe.”
Kyle Chan45,570 views • 3 months ago
Deng Xiaoping in 1978 in his heavy Sichuan accent: "The key to the Four Modernizations is the modernization of science and technology. Without modern science and technology, it is impossible to build modern agriculture, modern industry, and modern national defense. Without the rapid development of science and technology, there can be no rapid development of the national economy." 四个现代化，关键是科学技术的现代化。没有现代科学技术，就不可能建设现代农业、现代工业、现代国防。没有科学技术的高速度发展，也就不可能有国民经济的高速度发展。 From his famous speech at China's National Science conference. I wrote about this quote in my new piece: CHINA'S TECHNOLOGY LONG GAME
Kyle Chan51,614 views • 4 months ago
BYD’s Americas head said that BYD plans to build a factory in Mexico but stresses that this is for the Mexican market and not for export to the US. She said the US market is too complicated, including the politics, whereas in China if you don’t go electric, “you have no future.”
Kyle Chan85,042 views • 2 years ago
Here’s the part where Zhipu’s CEO talks about AI in 2027 reaching a potential dawn for AGI. Chinese and English transcript: Xiaojun: 包括大语言模型、图像模型和世界模型，需要把它们统一到一个更大的模型中，你认可他这个观点吗？他跟我们刚才说的是一个事对吧？ Including large language models, image models, and world models, they need to be unified into a larger model. Do you agree with his view? This is the same thing we were just talking about, right? Zhang Peng: 其实说的是一件事情。就不同模态和你不同的任务，怎么把它真正找到一个统一的这个建模的方式，原生地把它融合到一起，而不是用系统化的方法。 It's actually talking about the same thing. For different modalities and different tasks, how to truly find a unified modeling method to natively integrate them together, rather than using a systemic approach. Xiaojun: 如果这个实现了会是AGI吗？ If this is realized, will it be AGI? Zhang Peng: 我觉得AGI就看到曙光了应该。再加上刚才我说的那个在线学习，可能它就真的看到曙光。你可以想象这样嘛，首先第一，你造出来一个脑子，这个脑子呢，其实各种能力都有，语言能力、对图像的理解能力，然后对物理世界的这种判断能力、识别能力都有。然后再给它装上手脚，它能去叫世界模型去解决这个个问题嘛，它能去预测这个世界有发生什么事情，然后再跟世界进行交互。然后呢，交互的结果再反馈回来，变成一个强化的信号，然后我又立刻地、马上地接受这个强化学习的信号再学习，修改我的模型。这样闭环起来。那这样的话可能就是那个谁说的，Demis，对，你要选择让人工智能在什么范围什么什么时候开始授权，让它自己去探索这个世界。那个时候可能就近了。 I think we would see the dawn of AGI. Plus the online learning I just mentioned, maybe we would truly see the dawn. You can imagine it like this: first, you build a brain. This brain actually has all kinds of capabilities—language ability, image understanding ability, and the ability to judge and recognize the physical world. Then you equip it with hands and feet so it can call upon the world model to solve problems, predict what will happen in the world, and interact with the world. Then, the results of that interaction are fed back as a reinforcement signal. Then I immediately receive this reinforcement learning signal, learn again, and modify my model. This forms a closed loop. If that happens, it might be like what Demis said, yes, you have to choose in what scope and at what time to start authorizing artificial intelligence to explore the world on its own. At that point, it might be close. Xiaojun: 这你觉得还有多远啊？ How far away do you think this is? Zhang Peng: 我其实也说不好有多远。好像我看了好几种说法。一个是说的，可能2027年要开始有这个能力，达到我刚才说的那个状态，然后剩下就是等待，看它自己去学，学到什么程度，能不能学到比如说超过人或者什么之类，接近人或者超过人。接近人和超过人其实就是我们说的基本上AGI的这个目标就到了嘛。那就2027年到现在还有两年时间，然后2027年之后，可能还需要花几年时间去调整这个效率啊或者是学习的这个成果怎么安全啊等等这些事情。可能我理性地判断这件事情可能需要比如说五年、八年这样的时间。 I can't really say for sure how far it is. I seem to have seen several different predictions. One says that maybe by 2027 it will start to have this capability, reaching the state I just described, and the rest is just waiting, letting it learn on its own to see to what extent it can learn—whether it can, for example, surpass humans or something similar, approach humans or surpass humans. Approaching and surpassing humans basically means the goal of AGI has been reached. So from now to 2027 is two years, and then after 2027, it might take a few more years to adjust efficiency, ensure the safety of what it has learned, and so on. My rational judgment is that this process might take, say, five or eight years. Xiaojun: 你们都会跟进吗？如果这个是AGI的正确方向？ Will you follow this path? If this is the right direction for AGI? Zhang Peng: 一定会。我们永远不会忘记我们的目标就是AGI。 Absolutely. We will never forget that our goal is AGI.
Kyle Chan24,406 views • 5 months ago
China’s new sense of confidence, as captured by Xi’s words at the latest economic work conference: “Experience has proven that attempts to ‘choke us’ cannot succeed.” 实践证明，对我们“卡脖子”是卡不住的 In this video, you also see Xi famously saying “Key core technologies cannot be begged, bought, or requested.” 关键核心技术是要不来、买不来、讨不来的 (in 2018)
Kyle Chan19,175 views • 7 months ago
This is the new Xiaomi YU7 electric SUV's smart driving system tested on real Beijing streets. Impressive, but a few glitches: 1) right-turn signal turns on randomly 2) wrongly thinks it's about to hit a truck 3) thinks a green left-turn light is red and stops 4) to get around a stopped car, it waits till the last minute to change lanes 5) slows too early before a turn With the truck incident, a guy sitting in the back explains that the system is more sensitive around large vehicles, pedestrians, and non-motorized vehicles (like bikes). Keep in mind, it's only marketed as an assisted driving system, not a fully autonomous driving system. Here's the original video on Weibo (Chinese only):
Kyle Chan14,084 views • 1 year ago
No more content to load